The CoolMolecules database is ideal for helping students learn how to predict molecular shape from molecular or structural formulas.
We use "sigma structures" (as shown on the right) for this purpose, but you could use Lewis structures. Sigma structures are simply
the structures on the road to Lewis structures that have all electrons shown but no multiple bonds yet.
See the Construct a Lewis Structure web page for detailed, step-by-step information about writing sigma structures.
The activity described below involves the discovery-mode learning we call data-driven chemistry.
(See R. M. Hanson, S. A. Bergman, "Data-Driven Chemistry: Making Molecular Models (Literally) from Electron Diffraction Data," J. Chem. Educ. 1994, 150)
Using this method, students start with sorting the data into different categories. Only after
they start to see patterns in the data do they learn about the theory that explains those patterns. In this case, particularly using sigma structures,
students quickly see that there is a correlation between the
arrangement of bonds and lone pairs around the central atom
and the geometry around the central atom.